2009-03-20 03:26:15 +08:00
|
|
|
/*
|
2010-03-05 12:35:37 +08:00
|
|
|
* trace event based perf event profiling/tracing
|
2009-03-20 03:26:15 +08:00
|
|
|
*
|
|
|
|
* Copyright (C) 2009 Red Hat Inc, Peter Zijlstra <pzijlstr@redhat.com>
|
2010-03-03 14:16:16 +08:00
|
|
|
* Copyright (C) 2009-2010 Frederic Weisbecker <fweisbec@gmail.com>
|
2009-03-20 03:26:15 +08:00
|
|
|
*/
|
|
|
|
|
2009-08-24 12:19:47 +08:00
|
|
|
#include <linux/module.h>
|
2010-01-28 09:32:29 +08:00
|
|
|
#include <linux/kprobes.h>
|
2009-03-20 03:26:15 +08:00
|
|
|
#include "trace.h"
|
|
|
|
|
2010-08-11 11:47:59 +08:00
|
|
|
static char __percpu *perf_trace_buf[PERF_NR_CONTEXTS];
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2010-03-23 07:08:59 +08:00
|
|
|
/*
|
|
|
|
* Force it to be aligned to unsigned long to avoid misaligned accesses
|
|
|
|
* suprises
|
|
|
|
*/
|
|
|
|
typedef typeof(unsigned long [PERF_MAX_TRACE_SIZE / sizeof(unsigned long)])
|
|
|
|
perf_trace_t;
|
2009-11-22 12:26:55 +08:00
|
|
|
|
2009-09-18 12:10:28 +08:00
|
|
|
/* Count the events in use (per event id, not per instance) */
|
2010-03-05 12:35:37 +08:00
|
|
|
static int total_ref_count;
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2010-11-18 08:39:17 +08:00
|
|
|
static int perf_trace_event_perm(struct ftrace_event_call *tp_event,
|
|
|
|
struct perf_event *p_event)
|
|
|
|
{
|
2013-11-14 23:23:04 +08:00
|
|
|
if (tp_event->perf_perm) {
|
|
|
|
int ret = tp_event->perf_perm(tp_event, p_event);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2012-02-15 22:51:52 +08:00
|
|
|
/* The ftrace function trace is allowed only for root. */
|
2014-03-02 23:56:38 +08:00
|
|
|
if (ftrace_event_is_function(tp_event)) {
|
|
|
|
if (perf_paranoid_tracepoint_raw() && !capable(CAP_SYS_ADMIN))
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We don't allow user space callchains for function trace
|
|
|
|
* event, due to issues with page faults while tracing page
|
|
|
|
* fault handler and its overall trickiness nature.
|
|
|
|
*/
|
|
|
|
if (!p_event->attr.exclude_callchain_user)
|
|
|
|
return -EINVAL;
|
2014-03-02 23:56:39 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Same reason to disable user stack dump as for user space
|
|
|
|
* callchains above.
|
|
|
|
*/
|
|
|
|
if (p_event->attr.sample_type & PERF_SAMPLE_STACK_USER)
|
|
|
|
return -EINVAL;
|
2014-03-02 23:56:38 +08:00
|
|
|
}
|
2012-02-15 22:51:52 +08:00
|
|
|
|
2010-11-18 08:39:17 +08:00
|
|
|
/* No tracing, just counting, so no obvious leak */
|
|
|
|
if (!(p_event->attr.sample_type & PERF_SAMPLE_RAW))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Some events are ok to be traced by non-root users... */
|
|
|
|
if (p_event->attach_state == PERF_ATTACH_TASK) {
|
|
|
|
if (tp_event->flags & TRACE_EVENT_FL_CAP_ANY)
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* ...otherwise raw tracepoint data can be a severe data leak,
|
|
|
|
* only allow root to have these.
|
|
|
|
*/
|
|
|
|
if (perf_paranoid_tracepoint_raw() && !capable(CAP_SYS_ADMIN))
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-02-15 22:51:49 +08:00
|
|
|
static int perf_trace_event_reg(struct ftrace_event_call *tp_event,
|
|
|
|
struct perf_event *p_event)
|
2009-09-18 06:54:43 +08:00
|
|
|
{
|
2010-08-11 11:47:59 +08:00
|
|
|
struct hlist_head __percpu *list;
|
2012-02-15 22:51:49 +08:00
|
|
|
int ret = -ENOMEM;
|
2010-05-19 20:02:22 +08:00
|
|
|
int cpu;
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2010-05-19 20:02:22 +08:00
|
|
|
p_event->tp_event = tp_event;
|
|
|
|
if (tp_event->perf_refcount++ > 0)
|
2009-09-18 06:54:43 +08:00
|
|
|
return 0;
|
|
|
|
|
2010-05-19 20:02:22 +08:00
|
|
|
list = alloc_percpu(struct hlist_head);
|
|
|
|
if (!list)
|
|
|
|
goto fail;
|
|
|
|
|
|
|
|
for_each_possible_cpu(cpu)
|
|
|
|
INIT_HLIST_HEAD(per_cpu_ptr(list, cpu));
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2010-05-19 20:02:22 +08:00
|
|
|
tp_event->perf_events = list;
|
2009-09-18 06:54:43 +08:00
|
|
|
|
2010-03-05 12:35:37 +08:00
|
|
|
if (!total_ref_count) {
|
2010-08-11 11:47:59 +08:00
|
|
|
char __percpu *buf;
|
2010-05-19 16:52:27 +08:00
|
|
|
int i;
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2010-08-15 02:45:13 +08:00
|
|
|
for (i = 0; i < PERF_NR_CONTEXTS; i++) {
|
2010-08-11 11:47:59 +08:00
|
|
|
buf = (char __percpu *)alloc_percpu(perf_trace_t);
|
2010-05-19 16:52:27 +08:00
|
|
|
if (!buf)
|
2010-05-19 20:02:22 +08:00
|
|
|
goto fail;
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2010-05-19 20:02:22 +08:00
|
|
|
perf_trace_buf[i] = buf;
|
2010-05-19 16:52:27 +08:00
|
|
|
}
|
2009-09-18 12:10:28 +08:00
|
|
|
}
|
|
|
|
|
2012-02-15 22:51:49 +08:00
|
|
|
ret = tp_event->class->reg(tp_event, TRACE_REG_PERF_REGISTER, NULL);
|
2010-05-19 20:02:22 +08:00
|
|
|
if (ret)
|
|
|
|
goto fail;
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2010-05-19 20:02:22 +08:00
|
|
|
total_ref_count++;
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
fail:
|
2010-03-05 12:35:37 +08:00
|
|
|
if (!total_ref_count) {
|
2010-05-19 16:52:27 +08:00
|
|
|
int i;
|
|
|
|
|
2010-08-15 02:45:13 +08:00
|
|
|
for (i = 0; i < PERF_NR_CONTEXTS; i++) {
|
2010-05-19 16:52:27 +08:00
|
|
|
free_percpu(perf_trace_buf[i]);
|
|
|
|
perf_trace_buf[i] = NULL;
|
|
|
|
}
|
2009-10-03 20:55:18 +08:00
|
|
|
}
|
2010-05-19 20:02:22 +08:00
|
|
|
|
|
|
|
if (!--tp_event->perf_refcount) {
|
|
|
|
free_percpu(tp_event->perf_events);
|
|
|
|
tp_event->perf_events = NULL;
|
2009-10-03 20:55:18 +08:00
|
|
|
}
|
2009-09-18 12:10:28 +08:00
|
|
|
|
|
|
|
return ret;
|
2009-09-18 06:54:43 +08:00
|
|
|
}
|
|
|
|
|
2012-02-15 22:51:49 +08:00
|
|
|
static void perf_trace_event_unreg(struct perf_event *p_event)
|
|
|
|
{
|
|
|
|
struct ftrace_event_call *tp_event = p_event->tp_event;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (--tp_event->perf_refcount > 0)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
tp_event->class->reg(tp_event, TRACE_REG_PERF_UNREGISTER, NULL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Ensure our callback won't be called anymore. The buffers
|
|
|
|
* will be freed after that.
|
|
|
|
*/
|
|
|
|
tracepoint_synchronize_unregister();
|
|
|
|
|
|
|
|
free_percpu(tp_event->perf_events);
|
|
|
|
tp_event->perf_events = NULL;
|
|
|
|
|
|
|
|
if (!--total_ref_count) {
|
|
|
|
for (i = 0; i < PERF_NR_CONTEXTS; i++) {
|
|
|
|
free_percpu(perf_trace_buf[i]);
|
|
|
|
perf_trace_buf[i] = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
module_put(tp_event->mod);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int perf_trace_event_open(struct perf_event *p_event)
|
|
|
|
{
|
|
|
|
struct ftrace_event_call *tp_event = p_event->tp_event;
|
|
|
|
return tp_event->class->reg(tp_event, TRACE_REG_PERF_OPEN, p_event);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void perf_trace_event_close(struct perf_event *p_event)
|
|
|
|
{
|
|
|
|
struct ftrace_event_call *tp_event = p_event->tp_event;
|
|
|
|
tp_event->class->reg(tp_event, TRACE_REG_PERF_CLOSE, p_event);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int perf_trace_event_init(struct ftrace_event_call *tp_event,
|
|
|
|
struct perf_event *p_event)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = perf_trace_event_perm(tp_event, p_event);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = perf_trace_event_reg(tp_event, p_event);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = perf_trace_event_open(p_event);
|
|
|
|
if (ret) {
|
|
|
|
perf_trace_event_unreg(p_event);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-05-19 20:02:22 +08:00
|
|
|
int perf_trace_init(struct perf_event *p_event)
|
2009-03-20 03:26:15 +08:00
|
|
|
{
|
2010-05-19 20:02:22 +08:00
|
|
|
struct ftrace_event_call *tp_event;
|
2013-11-16 01:39:45 +08:00
|
|
|
u64 event_id = p_event->attr.config;
|
2009-05-06 10:33:45 +08:00
|
|
|
int ret = -EINVAL;
|
2009-03-20 03:26:15 +08:00
|
|
|
|
2009-05-06 10:33:45 +08:00
|
|
|
mutex_lock(&event_mutex);
|
2010-05-19 20:02:22 +08:00
|
|
|
list_for_each_entry(tp_event, &ftrace_events, list) {
|
2010-05-21 23:49:57 +08:00
|
|
|
if (tp_event->event.type == event_id &&
|
2010-06-08 23:22:06 +08:00
|
|
|
tp_event->class && tp_event->class->reg &&
|
2010-05-19 20:02:22 +08:00
|
|
|
try_module_get(tp_event->mod)) {
|
|
|
|
ret = perf_trace_event_init(tp_event, p_event);
|
2010-09-01 18:58:43 +08:00
|
|
|
if (ret)
|
|
|
|
module_put(tp_event->mod);
|
2009-05-06 10:33:45 +08:00
|
|
|
break;
|
|
|
|
}
|
2009-03-20 03:26:15 +08:00
|
|
|
}
|
2009-05-06 10:33:45 +08:00
|
|
|
mutex_unlock(&event_mutex);
|
2009-03-20 03:26:15 +08:00
|
|
|
|
2009-05-06 10:33:45 +08:00
|
|
|
return ret;
|
2009-03-20 03:26:15 +08:00
|
|
|
}
|
|
|
|
|
2012-02-15 22:51:49 +08:00
|
|
|
void perf_trace_destroy(struct perf_event *p_event)
|
|
|
|
{
|
|
|
|
mutex_lock(&event_mutex);
|
|
|
|
perf_trace_event_close(p_event);
|
|
|
|
perf_trace_event_unreg(p_event);
|
|
|
|
mutex_unlock(&event_mutex);
|
|
|
|
}
|
|
|
|
|
perf: Rework the PMU methods
Replace pmu::{enable,disable,start,stop,unthrottle} with
pmu::{add,del,start,stop}, all of which take a flags argument.
The new interface extends the capability to stop a counter while
keeping it scheduled on the PMU. We replace the throttled state with
the generic stopped state.
This also allows us to efficiently stop/start counters over certain
code paths (like IRQ handlers).
It also allows scheduling a counter without it starting, allowing for
a generic frozen state (useful for rotating stopped counters).
The stopped state is implemented in two different ways, depending on
how the architecture implemented the throttled state:
1) We disable the counter:
a) the pmu has per-counter enable bits, we flip that
b) we program a NOP event, preserving the counter state
2) We store the counter state and ignore all read/overflow events
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: paulus <paulus@samba.org>
Cc: stephane eranian <eranian@googlemail.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Lin Ming <ming.m.lin@intel.com>
Cc: Yanmin <yanmin_zhang@linux.intel.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: David Miller <davem@davemloft.net>
Cc: Michael Cree <mcree@orcon.net.nz>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-06-16 20:37:10 +08:00
|
|
|
int perf_trace_add(struct perf_event *p_event, int flags)
|
2009-09-18 06:54:43 +08:00
|
|
|
{
|
2010-05-19 20:02:22 +08:00
|
|
|
struct ftrace_event_call *tp_event = p_event->tp_event;
|
2010-08-11 11:47:59 +08:00
|
|
|
struct hlist_head __percpu *pcpu_list;
|
2010-05-19 20:02:22 +08:00
|
|
|
struct hlist_head *list;
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2010-08-11 11:47:59 +08:00
|
|
|
pcpu_list = tp_event->perf_events;
|
|
|
|
if (WARN_ON_ONCE(!pcpu_list))
|
2010-05-19 20:02:22 +08:00
|
|
|
return -EINVAL;
|
2009-09-18 12:10:28 +08:00
|
|
|
|
perf: Rework the PMU methods
Replace pmu::{enable,disable,start,stop,unthrottle} with
pmu::{add,del,start,stop}, all of which take a flags argument.
The new interface extends the capability to stop a counter while
keeping it scheduled on the PMU. We replace the throttled state with
the generic stopped state.
This also allows us to efficiently stop/start counters over certain
code paths (like IRQ handlers).
It also allows scheduling a counter without it starting, allowing for
a generic frozen state (useful for rotating stopped counters).
The stopped state is implemented in two different ways, depending on
how the architecture implemented the throttled state:
1) We disable the counter:
a) the pmu has per-counter enable bits, we flip that
b) we program a NOP event, preserving the counter state
2) We store the counter state and ignore all read/overflow events
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: paulus <paulus@samba.org>
Cc: stephane eranian <eranian@googlemail.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Lin Ming <ming.m.lin@intel.com>
Cc: Yanmin <yanmin_zhang@linux.intel.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: David Miller <davem@davemloft.net>
Cc: Michael Cree <mcree@orcon.net.nz>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-06-16 20:37:10 +08:00
|
|
|
if (!(flags & PERF_EF_START))
|
|
|
|
p_event->hw.state = PERF_HES_STOPPED;
|
|
|
|
|
2010-08-11 11:47:59 +08:00
|
|
|
list = this_cpu_ptr(pcpu_list);
|
2010-05-19 20:02:22 +08:00
|
|
|
hlist_add_head_rcu(&p_event->hlist_entry, list);
|
2009-09-18 12:10:28 +08:00
|
|
|
|
2012-02-15 22:51:50 +08:00
|
|
|
return tp_event->class->reg(tp_event, TRACE_REG_PERF_ADD, p_event);
|
2010-05-19 20:02:22 +08:00
|
|
|
}
|
2009-09-18 12:10:28 +08:00
|
|
|
|
perf: Rework the PMU methods
Replace pmu::{enable,disable,start,stop,unthrottle} with
pmu::{add,del,start,stop}, all of which take a flags argument.
The new interface extends the capability to stop a counter while
keeping it scheduled on the PMU. We replace the throttled state with
the generic stopped state.
This also allows us to efficiently stop/start counters over certain
code paths (like IRQ handlers).
It also allows scheduling a counter without it starting, allowing for
a generic frozen state (useful for rotating stopped counters).
The stopped state is implemented in two different ways, depending on
how the architecture implemented the throttled state:
1) We disable the counter:
a) the pmu has per-counter enable bits, we flip that
b) we program a NOP event, preserving the counter state
2) We store the counter state and ignore all read/overflow events
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: paulus <paulus@samba.org>
Cc: stephane eranian <eranian@googlemail.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Lin Ming <ming.m.lin@intel.com>
Cc: Yanmin <yanmin_zhang@linux.intel.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: David Miller <davem@davemloft.net>
Cc: Michael Cree <mcree@orcon.net.nz>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-06-16 20:37:10 +08:00
|
|
|
void perf_trace_del(struct perf_event *p_event, int flags)
|
2010-05-19 20:02:22 +08:00
|
|
|
{
|
2012-02-15 22:51:50 +08:00
|
|
|
struct ftrace_event_call *tp_event = p_event->tp_event;
|
2010-05-19 20:02:22 +08:00
|
|
|
hlist_del_rcu(&p_event->hlist_entry);
|
2012-02-15 22:51:50 +08:00
|
|
|
tp_event->class->reg(tp_event, TRACE_REG_PERF_DEL, p_event);
|
2009-09-18 06:54:43 +08:00
|
|
|
}
|
|
|
|
|
2010-03-05 12:35:37 +08:00
|
|
|
__kprobes void *perf_trace_buf_prepare(int size, unsigned short type,
|
2010-05-19 16:52:27 +08:00
|
|
|
struct pt_regs *regs, int *rctxp)
|
2010-01-28 09:32:29 +08:00
|
|
|
{
|
|
|
|
struct trace_entry *entry;
|
2010-05-25 17:02:55 +08:00
|
|
|
unsigned long flags;
|
2010-05-19 20:02:22 +08:00
|
|
|
char *raw_data;
|
2010-05-19 16:52:27 +08:00
|
|
|
int pc;
|
2010-01-28 09:32:29 +08:00
|
|
|
|
2010-03-23 07:08:59 +08:00
|
|
|
BUILD_BUG_ON(PERF_MAX_TRACE_SIZE % sizeof(unsigned long));
|
|
|
|
|
2013-06-18 01:02:11 +08:00
|
|
|
if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE,
|
|
|
|
"perf buffer not large enough"))
|
|
|
|
return NULL;
|
|
|
|
|
2010-01-28 09:32:29 +08:00
|
|
|
pc = preempt_count();
|
|
|
|
|
|
|
|
*rctxp = perf_swevent_get_recursion_context();
|
|
|
|
if (*rctxp < 0)
|
2010-05-19 20:02:22 +08:00
|
|
|
return NULL;
|
2010-01-28 09:32:29 +08:00
|
|
|
|
2010-05-21 18:31:09 +08:00
|
|
|
raw_data = this_cpu_ptr(perf_trace_buf[*rctxp]);
|
2010-01-28 09:32:29 +08:00
|
|
|
|
|
|
|
/* zero the dead bytes from align to not leak stack to user */
|
2010-03-23 07:08:59 +08:00
|
|
|
memset(&raw_data[size - sizeof(u64)], 0, sizeof(u64));
|
2010-01-28 09:32:29 +08:00
|
|
|
|
|
|
|
entry = (struct trace_entry *)raw_data;
|
2010-05-25 17:02:55 +08:00
|
|
|
local_save_flags(flags);
|
|
|
|
tracing_generic_entry_update(entry, flags, pc);
|
2010-01-28 09:32:29 +08:00
|
|
|
entry->type = type;
|
|
|
|
|
|
|
|
return raw_data;
|
|
|
|
}
|
2010-03-05 12:35:37 +08:00
|
|
|
EXPORT_SYMBOL_GPL(perf_trace_buf_prepare);
|
2012-02-15 22:51:52 +08:00
|
|
|
|
|
|
|
#ifdef CONFIG_FUNCTION_TRACER
|
|
|
|
static void
|
2011-08-09 04:57:47 +08:00
|
|
|
perf_ftrace_function_call(unsigned long ip, unsigned long parent_ip,
|
2011-08-10 00:50:46 +08:00
|
|
|
struct ftrace_ops *ops, struct pt_regs *pt_regs)
|
2012-02-15 22:51:52 +08:00
|
|
|
{
|
|
|
|
struct ftrace_entry *entry;
|
|
|
|
struct hlist_head *head;
|
|
|
|
struct pt_regs regs;
|
|
|
|
int rctx;
|
|
|
|
|
2013-06-18 01:02:04 +08:00
|
|
|
head = this_cpu_ptr(event_function.perf_events);
|
|
|
|
if (hlist_empty(head))
|
|
|
|
return;
|
|
|
|
|
2012-02-15 22:51:52 +08:00
|
|
|
#define ENTRY_SIZE (ALIGN(sizeof(struct ftrace_entry) + sizeof(u32), \
|
|
|
|
sizeof(u64)) - sizeof(u32))
|
|
|
|
|
|
|
|
BUILD_BUG_ON(ENTRY_SIZE > PERF_MAX_TRACE_SIZE);
|
|
|
|
|
|
|
|
perf_fetch_caller_regs(®s);
|
|
|
|
|
|
|
|
entry = perf_trace_buf_prepare(ENTRY_SIZE, TRACE_FN, NULL, &rctx);
|
|
|
|
if (!entry)
|
|
|
|
return;
|
|
|
|
|
|
|
|
entry->ip = ip;
|
|
|
|
entry->parent_ip = parent_ip;
|
|
|
|
perf_trace_buf_submit(entry, ENTRY_SIZE, rctx, 0,
|
2012-07-11 22:14:58 +08:00
|
|
|
1, ®s, head, NULL);
|
2012-02-15 22:51:52 +08:00
|
|
|
|
|
|
|
#undef ENTRY_SIZE
|
|
|
|
}
|
|
|
|
|
|
|
|
static int perf_ftrace_function_register(struct perf_event *event)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *ops = &event->ftrace_ops;
|
|
|
|
|
|
|
|
ops->flags |= FTRACE_OPS_FL_CONTROL;
|
|
|
|
ops->func = perf_ftrace_function_call;
|
|
|
|
return register_ftrace_function(ops);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int perf_ftrace_function_unregister(struct perf_event *event)
|
|
|
|
{
|
|
|
|
struct ftrace_ops *ops = &event->ftrace_ops;
|
ftrace, perf: Add filter support for function trace event
Adding support to filter function trace event via perf
interface. It is now possible to use filter interface
in the perf tool like:
perf record -e ftrace:function --filter="(ip == mm_*)" ls
The filter syntax is restricted to the the 'ip' field only,
and following operators are accepted '==' '!=' '||', ending
up with the filter strings like:
ip == f1[, ]f2 ... || ip != f3[, ]f4 ...
with comma ',' or space ' ' as a function separator. If the
space ' ' is used as a separator, the right side of the
assignment needs to be enclosed in double quotes '"', e.g.:
perf record -e ftrace:function --filter '(ip == do_execve,sys_*,ext*)' ls
perf record -e ftrace:function --filter '(ip == "do_execve,sys_*,ext*")' ls
perf record -e ftrace:function --filter '(ip == "do_execve sys_* ext*")' ls
The '==' operator adds trace filter with same effect as would
be added via set_ftrace_filter file.
The '!=' operator adds trace filter with same effect as would
be added via set_ftrace_notrace file.
The right side of the '!=', '==' operators is list of functions
or regexp. to be added to filter separated by space.
The '||' operator is used for connecting multiple filter definitions
together. It is possible to have more than one '==' and '!='
operators within one filter string.
Link: http://lkml.kernel.org/r/1329317514-8131-8-git-send-email-jolsa@redhat.com
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2012-02-15 22:51:54 +08:00
|
|
|
int ret = unregister_ftrace_function(ops);
|
|
|
|
ftrace_free_filter(ops);
|
|
|
|
return ret;
|
2012-02-15 22:51:52 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void perf_ftrace_function_enable(struct perf_event *event)
|
|
|
|
{
|
|
|
|
ftrace_function_local_enable(&event->ftrace_ops);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void perf_ftrace_function_disable(struct perf_event *event)
|
|
|
|
{
|
|
|
|
ftrace_function_local_disable(&event->ftrace_ops);
|
|
|
|
}
|
|
|
|
|
|
|
|
int perf_ftrace_event_register(struct ftrace_event_call *call,
|
|
|
|
enum trace_reg type, void *data)
|
|
|
|
{
|
|
|
|
switch (type) {
|
|
|
|
case TRACE_REG_REGISTER:
|
|
|
|
case TRACE_REG_UNREGISTER:
|
|
|
|
break;
|
|
|
|
case TRACE_REG_PERF_REGISTER:
|
|
|
|
case TRACE_REG_PERF_UNREGISTER:
|
|
|
|
return 0;
|
|
|
|
case TRACE_REG_PERF_OPEN:
|
|
|
|
return perf_ftrace_function_register(data);
|
|
|
|
case TRACE_REG_PERF_CLOSE:
|
|
|
|
return perf_ftrace_function_unregister(data);
|
|
|
|
case TRACE_REG_PERF_ADD:
|
|
|
|
perf_ftrace_function_enable(data);
|
|
|
|
return 0;
|
|
|
|
case TRACE_REG_PERF_DEL:
|
|
|
|
perf_ftrace_function_disable(data);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_FUNCTION_TRACER */
|